基于机器学习的新能源汽车残值评估方法

Research on Residual Value Evaluation Method of New Energy Vehicle Based on Machine Learning

针对现有汽车残值评估方法信息使用较少、严重依赖人工检测、误差大和无法解决新能源汽车残值评估的问题,基于梯度提升回归树模型,使用行驶里程、使用时间、功率、过户次数等多维特征数据,训练模型对残值进行预测。从选取惩罚项、树的深度、基学习器类型以及提取特征重要性方面优化模型。最后,使用Stacking模型集成算法对二阶多项式、XGBoost、LightGBM模型进行集成。实验结果表明,使用Stacking集成后的模型可以根据当前车况数据自动计算残值,不需要人工检测,具有实时性,较其他方法有更高的准确度。

Aiming at the problem of less use of information , heavy reliance on manual detection , large error and inability to solve the problem of residual value evaluation of new energy vehicles in the existing vehicle residual value assessment method , a method for estimating the residual value of new energy vehicles based on machine learning is proposed.The method , based on the gradient boosting regression tree model , uses the multi - dimensional feature data such as mileage , time , power , and number of transfers , and the training model to predict the residual value.The model is optimized from the selection of penalty terms , the depth of the tree , the type of base learner , and the extraction of feature importance information.Finally , the second - order polynomial , XGBoost , and LightGBM models are integrated using a stacked model integration algorithm.The experimental results show that the model with stack integration can automatically calculate the residual value according to the current vehicle condition data , without manual detection , and has real - time performance , which has higher accuracy than other methods.